5D Library › Power-Law Weighting and Bottom-Up Controllability in Multi-Tier Conditional Excitation Systems · 中文

Chapter 3 Endowment and Timing: The Published Evidence

Two tiers that are easily overrated

In the author’s pyramid the top two tiers are the endogenous prior C1C_1 and timing C2C_2, together weighted at 75 %. They are also the tiers most readily called “fate”: no one chooses how they are born or when. This chapter asks what published research says about them, and what it does not say. One point must be made at the outset: research measures only a small measurable part of each tier, such as genetic variants associated with educational attainment or the unemployment rate in the year of graduation. Neither equals the whole tier in the author’s sense.

How much can a polygenic score explain?

Evidence

Genome-wide association studies of educational attainment. In about 1.1 million people, Lee and colleagues found 1,271 independent genome-wide significant SNPs; a polygenic score built from the results explains 11–13 % of the variance in educational attainment in independent samples and 7–10 % of the variance in cognitive test performance (Lee et al. 2018). Okbay and colleagues extended the sample to about 3 million people and found 3,952 approximately uncorrelated significant SNPs; the score explains 12–16 % of the variance, but in within-family analyses (using sibling and parent–offspring data and controlling for the parents’ scores) direct genetic effects account for only about half the magnitude of the score’s association with attainment (Okbay et al. 2022).

Genetic nurture. Using Icelandic parent–offspring genotypes (21,637 probands with at least one genotyped parent), Kong and colleagues found that a score built from the alleles parents did not transmit to their child is also associated with the child’s educational attainment, at about 29.9 % of the effect of the transmitted score. Untransmitted genes can act only through the environment the parents create; the authors called this “genetic nurture” (Kong et al. 2018).

Portability. Martin and colleagues pointed out that current polygenic scores are mostly developed in samples of European ancestry and predict markedly less well in other ancestry groups (Martin et al. 2019).

Together these findings correct claims such as “endowment is half of it”. First, 11–16 % is predictive power at the population level and is weak for individuals: a score explaining 13 % of the variance leaves 87 % of the variance among people with the same score. Second, the figure is not all “endowment”: it mixes in genetic nurture (parents’ genes acting through the home), population stratification and assortative mating. Within-family comparisons remove these, and the direct effect that remains is about half the magnitude of the association, or roughly a quarter in terms of variance. Third, the figure depends on population and period: it changes in other ancestry groups and under other education systems.

Calculation

The program simulates 20,000 invented families with two children each. Each child’s score has variance 1; siblings share the parental mean (variance 0.5) and each has a segregation component (variance 0.5). Outcome = direct effect × own score + (genetic nurture + stratification) × parental mean + family environment + noise. The parameters are illustrative values chosen to give a population R2R^2 of about 13 %; they are not estimates.

# Why a polygenic score "explains" less within families than in the population (Ch. 3).
# Sibling pairs: each score g has variance 1; siblings share the parental mean m (variance
# 0.5) plus their own segregation term (variance 0.5). Outcome y = a g + (b + c) m +
# family environment + noise: a = direct effect, b = "genetic nurture", c = stratification
# travelling with the score. a, b, c are illustrative, chosen so that the population R^2
# is about 13 % (Lee et al. 2018 report 11-13 %); they are NOT estimates.
import math, random
from statistics import mean

def slope_r2(x, y):
    mx, my = mean(x), mean(y)
    sxy = sum((a - mx) * (b - my) for a, b in zip(x, y))
    sxx, syy = sum((a - mx) ** 2 for a in x), sum((b - my) ** 2 for b in y)
    return sxy / sxx, sxy ** 2 / (sxx * syy)

def families(n, a, b, c, port=1.0, seed=41):
    # port < 1: the score tags the causal variants less well (e.g. another ancestry)
    rng = random.Random(seed); g1, g2, y1, y2 = [], [], [], []
    for _ in range(n):
        m = rng.gauss(0, math.sqrt(0.5)); fam = rng.gauss(0, 0.45)
        for G, Y in ((g1, y1), (g2, y2)):
            g = m + rng.gauss(0, math.sqrt(0.5))
            Y.append(a * g + (b + c) * m + fam + rng.gauss(0, 0.80))
            G.append(port * g + math.sqrt(1 - port ** 2) * rng.gauss(0, 1))
    return g1, g2, y1, y2

if __name__ == "__main__":
    a, b, c = 0.18, 0.20, 0.16
    g1, g2, y1, y2 = families(20000, a, b, c)
    pop_b, pop_r2 = slope_r2(g1 + g2, y1 + y2)
    wf_b = slope_r2([p - q for p, q in zip(g1, g2)], [p - q for p, q in zip(y1, y2)])[0]
    my = mean(y1); vy = mean([(v - my) ** 2 for v in y1])
    print("20,000 simulated sibling pairs (illustrative parameters)")
    print(f"  population: slope {pop_b:.3f}, R^2 = {pop_r2:.1%}")
    print(f"  within sibling pairs: slope {wf_b:.3f} = direct effect only"
          f" (true a = {a}); ratio {wf_b / pop_b:.2f}")
    print(f"  variance due to the direct effect alone: a^2 / Var(y) = {a ** 2 / vy:.1%}")
    print(f"  expected population slope a + (b + c) * 0.5 = {a + (b + c) * 0.5:.3f}")
    for port in (0.8, 0.5):
        h1, h2, z1, z2 = families(20000, a, b, c, port=port, seed=42)
        print(f"  score tags causal variants with correlation {port}: population R^2 ="
              f" {slope_r2(h1 + h2, z1 + z2)[1]:.1%}")
# 20,000 simulated sibling pairs (illustrative parameters)
#   population: slope 0.363, R^2 = 13.2%
#   within sibling pairs: slope 0.180 = direct effect only (true a = 0.18); ratio 0.50
#   variance due to the direct effect alone: a^2 / Var(y) = 3.2%
#   expected population slope a + (b + c) * 0.5 = 0.360
#   score tags causal variants with correlation 0.8: population R^2 = 8.5%
#   score tags causal variants with correlation 0.5: population R^2 = 3.5%

In the population the slope of the outcome on the score is 0.363 and R2R^2 is 13.2 %. Differencing between siblings removes everything they share (parental mean, family environment); the slope falls to 0.180, exactly the direct effect that was set, half the population slope. The direct effect alone explains only 3.2 % of the variance, about a quarter of the population R2R^2. If the score tags the causal variants less well (correlation 0.8 or 0.5, as outside the discovery population), the population R2R^2 falls to 8.5 % and 3.5 %.

Note

These numbers cannot be used to judge anyone. Polygenic scores are tools for population research. The trial in this book (Chapter 8) collects no genetic data. Using scores for admissions, hiring or partner choice is scientifically unsound (weak individual prediction, poor portability) and ethically dangerous. In the author’s model the “endowment” tier also includes temperament, health and early family conditions, none of which can be summed up by a single score either.

Graduating into a recession

The most easily measured part of the timing tier is the state of the economy when one enters the labour market.

Evidence

US college graduates. Using white male college graduates of 1979–1989 in the US National Longitudinal Survey of Youth (NLSY79), with year of birth among the instruments, Kahn estimated the effect of economic conditions at graduation on nearly two decades of career: each percentage point of national unemployment at graduation lowers initial wages by about 6–7 %; the loss shrinks over time but is still statistically significant 15 years later (Kahn 2010).

Canadian college graduates. Using administrative data linking Canadian university graduates of 1976–1995 to tax records, Oreopoulos, von Wachter and Heisz found that a typical recession (a rise in unemployment of about 5 percentage points) implies an initial earnings loss of about 9 % that halves within five years and fades by ten; part of the loss comes from starting at lower-paying employers and is recovered gradually by moving to better firms (Oreopoulos et al. 2012).

Both studies show that timing has real and lasting effects. They also illustrate a distinction that is easily overlooked: an effect can be large for those affected without accounting for a large share of the differences across everyone.

Calculation

The program draws two illustrative loss paths: path A follows the summary of Oreopoulos and colleagues (1.8 % per percentage point at first, halving in five years, gone in ten); path B is a more persistent illustrative path (Kahn’s initial size, 6.5 % per point, halving every ten years). Assuming that unemployment at entry varies across graduation cohorts with a standard deviation of 1.5 points and that individual ten-year average log earnings have a standard deviation of 0.50 (both illustrative), it computes the share of the variance in earnings due to entry timing.

# Timing (layer C2): graduating into a recession (Chapter 3).
# Path A follows the summary in Oreopoulos, von Wachter & Heisz (2012): a typical
# recession (+5 percentage points of unemployment) means an initial earnings loss of about
# 9 % that halves within 5 years and fades to 0 by 10 years (drawn as straight lines).
# Path B is an illustrative, more persistent path with Kahn's (2010) initial size for US
# college graduates (6.5 % per point, the middle of her 6-7 %), halving every 10 years.
# Question: how large is the loss for the unlucky, and what share of the cross-sectional
# variance of 10-year average log earnings does entry timing explain? The cohort SD of
# entry unemployment (1.5 points) and the individual SD (0.50) are illustrative.
import math

def path_a(t, pts=5.0):
    per_pt = 9.0 / 5.0
    first = per_pt * pts
    return first - (first / 2) * t / 5 if t <= 5 else max(0.0, first / 2 * (10 - t) / 5)

def path_b(t, pts=5.0):
    return 6.5 * pts * 0.5 ** (t / 10)

def average_loss(path, years=10, pts=5.0, steps=1000):
    h = years / steps
    return sum(path((i + 0.5) * h, pts) for i in range(steps)) * h / years

if __name__ == "__main__":
    print("earnings loss (%) by years since entry")
    print("  years since entry:         " + " ".join(f"{t:5d}" for t in (0, 2, 5, 8, 10)))
    for name, p, pts in (("path A, +5 points", path_a, 5.0),
                         ("path A, +1 point ", path_a, 1.0),
                         ("path B, +1 point ", path_b, 1.0)):
        print(f"  {name}:  " + " ".join(f"{p(t, pts):5.1f}" for t in (0, 2, 5, 8, 10))
              + f"   10-yr mean {average_loss(p, pts=pts):.1f} %")
    sd_cohort, sd_person = 1.5, 0.50
    print(f"cohort SD of entry unemployment {sd_cohort} points;"
          f" individual SD {sd_person}")
    for name, p in (("path A", path_a), ("path B", path_b)):
        per_point = average_loss(p, pts=1.0) / 100
        share = (per_point * sd_cohort) ** 2 / sd_person ** 2
        print(f"  {name}: {per_point:.2%} of 10-year earnings per point;"
              f" variance share of timing {share:.2%}")
    print("a loss of several per cent for a whole cohort can be a tiny variance share")
# earnings loss (%) by years since entry
#   years since entry:             0     2     5     8    10
#   path A, +5 points:    9.0   7.2   4.5   1.8   0.0   10-yr mean 4.5 %
#   path A, +1 point :    1.8   1.4   0.9   0.4   0.0   10-yr mean 0.9 %
#   path B, +1 point :    6.5   5.7   4.6   3.7   3.2   10-yr mean 4.7 %
# cohort SD of entry unemployment 1.5 points; individual SD 0.5
#   path A: 0.90% of 10-year earnings per point; variance share of timing 0.07%
#   path B: 4.69% of 10-year earnings per point; variance share of timing 1.98%
# a loss of several per cent for a whole cohort can be a tiny variance share

In a 5-point recession, path A’s ten-year average loss is 4.5 %. Yet even on the more persistent path B, entry timing explains only about 2 % of the variance, and on path A only 0.07 %. The reason is simple: unemployment differs little between cohorts, while earnings differ a great deal within a cohort.

This result supports neither “timing is 25 %” nor “timing does not matter”. First, the national unemployment rate is only a small measurable part of the timing tier; the author’s “timing” also covers industry cycles and the windows of opportunity an individual meets, which are harder to measure and may matter more. Second, a variance share is not a measure of importance to the individual: for the cohort that graduated into a recession, a long-run earnings loss of several per cent is very real. Confusing the two is one of the most common errors in debates about “what matters most”. Volume 37 discussed the structure of cycles themselves, and it is not repeated here.

Hypothesis · not established

The timing-window hypothesis (the author’s hypothesis, not established). Individual-level “timing” (entering an industry during an expansion, being ready when an opportunity appears) affects career outcomes far more than the variance share of macroeconomic entry conditions suggests. Testable prediction: within one cohort and one industry, a pre-defined “window of opportunity” measure (say, the growth of one’s department in the two years after joining) explains clearly more of the variance of ten-year earnings than the unemployment rate in the year of entry (for example, more than 5 %). If individual window measures also explain only a tiny share, the hypothesis is weakened.

Chapter summary

Polygenic scores for educational attainment explain 11–16 % of the variance in the population, but the figure includes genetic nurture and stratification; within families the direct effect is about half the magnitude of the association and about a quarter in terms of variance; the scores predict markedly less well in other ancestry groups and cannot be used to judge individuals. Graduating into a recession brings real and lasting earnings losses, yet entry timing may account for only a small share of the variance in earnings across everyone. The evidence on neither tier supports “endowment and timing together are 75 %”, and neither supports “they do not matter”. The next chapter turns to environment, reputation and luck.

References

Kahn, Lisa B. 2010. “The Long-Term Labor Market Consequences of Graduating from College in a Bad Economy.” Labour Economics 17 (2): 303–16. https://doi.org/10.1016/j.labeco.2009.09.002.
Kong, Augustine, Gudmar Thorleifsson, Michael L. Frigge, et al. 2018. “The Nature of Nurture: Effects of Parental Genotypes.” Science 359 (6374): 424–28. https://doi.org/10.1126/science.aan6877.
Lee, James J., Robbee Wedow, Aysu Okbay, Edward Kong, et al. 2018. “Gene Discovery and Polygenic Prediction from a Genome-Wide Association Study of Educational Attainment in 1.1 Million Individuals.” Nature Genetics 50 (8): 1112–21. https://doi.org/10.1038/s41588-018-0147-3.
Martin, Alicia R., Masahiro Kanai, Yoichiro Kamatani, Yukinori Okada, Benjamin M. Neale, and Mark J. Daly. 2019. “Clinical Use of Current Polygenic Risk Scores May Exacerbate Health Disparities.” Nature Genetics 51 (4): 584–91. https://doi.org/10.1038/s41588-019-0379-x.
Okbay, Aysu, Yeda Wu, Nancy Wang, et al. 2022. “Polygenic Prediction of Educational Attainment Within and Between Families from Genome-Wide Association Analyses in 3 Million Individuals.” Nature Genetics 54 (4): 437–49. https://doi.org/10.1038/s41588-022-01016-z.
Oreopoulos, Philip, Till von Wachter, and Andrew Heisz. 2012. “The Short- and Long-Term Career Effects of Graduating in a Recession.” American Economic Journal: Applied Economics 4 (1): 1–29. https://doi.org/10.1257/app.4.1.1.

Fan, Y. (2026). Power-Law Weighting and Bottom-Up Controllability in Multi-Tier Conditional Excitation Systems: A Mathematical Synthesis of Endogenous Priors, Timing Windows, Network Topology and Reputational Games [多层条件激发系统中的幂律权重与逆向控制可及性] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23810093 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).